Teacher Specialization and Production of Human Capital
In a field experiment in Houston public schools, after two years students with specialized teachers lagged about a month behind those with non-specialized teachers.
Industrial production methods utilizing highly specialized labor have raised the quality of manufactured goods while substantially reducing their costs. This success has encouraged school reform advocates to suggest that schools might benefit by shifting to specialized teachers. The advocates contend that specializing in a narrow subject range would reduce teacher workload and boost student achievement because teachers would have more time to master their subject and to focus on lesson planning.
In The 'Pupil' Factory: Specialization and the Production of Human Capital in Schools (NBER Working Paper No. 22205), Roland G. Fryer, Jr. reports on results of a randomized field experiment testing that theory. At the end of two school years that began in 2013–14, students in Houston public elementary schools with specialized teachers lagged approximately one month behind students in schools with non-specialized teachers. The lags for students with special needs and for students with younger teachers were considerably larger. Students in schools with specialized teachers also had lower attendance and a higher probability of suspension due to poor behavior.
The sample included 50 schools. Half of the schools used specialized teachers. Half were control schools, which maintained the status quo. Students in the schools that were included in the experiment, both with and without specialized teachers, were less likely to be white or Asian, more likely to be economically disadvantaged, and had lower test scores in math and reading than in other schools in the city of Houston. According to Fryer, "the results estimated are likely to be more applicable to urban schools with higher concentrations of minority students."
The teacher specialties were math, science, social studies, and reading. Principals assigned teachers to subjects based on average teacher value-added measures, classroom observation, or recommendations. Specialized teachers were not allowed to teach both math and reading. Other combinations were allowed. The most common division of labor was a two-teacher team in which one teacher taught reading and social studies and the other teacher taught math and science. Although students had different teachers for different subjects, they stayed with the same group of classmates for all subjects.
Data on individual students included standard demographic variables, behavior and attendance records, English proficiency, state math and reading test scores, and nationally normed Stanford 10 subject scores in math and reading. Schools were randomized by ranking them by the sum of their mean reading and math test scores in the previous two years. They were grouped into pairs by going down the list in order. Then the schools in each pair were randomly assigned to the specialized or to the control group.
Different specifications did not change the negative result for specialization. It was robust to several sensitivity tests. It remained negative whether high- or low-stakes test scores were used to measure outcomes, was unchanged when tested for the effects of student attrition from specific schools, and apparently was not the result of unknown differences in individual schools.
According to data from a teacher survey, teachers in specialized schools were less likely to report providing students with individual attention. Because specialized teachers teach more students, they may find it more difficult to get to know each student and to tailor lessons to the needs of individual students in a particular class. As a result, the gains from specialization may have been outweighed by the losses caused by inefficient teaching choices for specific students. The author concludes by emphasizing that these results "provide a cautionary tale about the potential productivity benefits of the division of labor when applied to human capital development."
The Digest is not copyrighted and may be reproduced freely with appropriate attribution of source.